[webkit-dev] Implementing WebTiming as a part of HTML5

[Alois Reitbauer]

As discussed with Richard and Zhiheng here the feedback on the proposal
by dynaTrace software. We had a look at the new spec draft and have a
number of questions and proposals:

 -    In the NavigationTiming interface you refer to paint events. Do
you really mean paint here or layouting? Paint events are in our opinion
less relevant from a performance perspective than layouting (or reflow)
events.

-     Additionally it would be nice to not only see the pain event count
but also the paint event time. You can then find out how much time is
spent layouting. Just having the number of operations does not provide
this information.

-     We are a bit unclear how to access the timing information for the
HTML elements. Would we have to browse the whole DOM to get them? This
could result in quite a performance overhead. 

-     Additionally we wonder if you only want to additionally support
timings for XHR requests. We think it would be good to support every
type of network request. 

-     We are also wondering about what to use the TimingMap respectively
the ticks property for.  A nice use case would be to store the time of a
click event and then retrieve it on the next page. However this will not
be part of the window.timing attribute - right?

On the mailing there was also a note that "... some parts are
implemented by chromium".  Are there any more details which parts, so we
could try to use them and provide feedback as well.

 

- Alois

 

Alois Reitbauer, Technology Strategist, dynaTrace software GmbH 



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-dev/attachments/20100215/e054f1b9/attachment.html>